An AI Readiness Framework an Owner Can Score

An AI readiness framework is a way to score the business before you score a tool. We use it when an owner wants a plan and all anyone has offered is a product tour. The tour can wait, because the score is about the work.

Cited in this note

63%

NIST groups AI risk work as Govern, Map, Measure, and Manage.¹ This note uses four groups an owner can see, which are data, process, people, and the decision. We call each group absent, partial, in place, or running. Those words are anchors and not a percentage. If you cannot point at the thing, you do not get the stronger word. We would rather call data absent than call it running because the owner feels modern.

The readiness guide is the matching page. The assessment is the longer note, and the questionnaire is the list of questions. The framework is what you do with the answers once they are on paper. You do not turn them into a fake share of readiness. You give each group a word you can defend.

What the bands mean

The Readiness Check scores 0 to 39 as Early, 40 to 59 as Forming, 60 to 79 as Ready, and 80 to 100 as Running.² The check helps you see where to look next. It does not certify a business.

Read those bands as a shape and not as a medal. Early means the work that matters still lives in people’s heads. Forming means some of it is written down, but nobody clearly owns it. Ready means you could put a system on the job without making up the job at the same time. Running means the tools already do real work, and the leftover pain is the task everyone still hands to one person. A band without those sentences is just a number. We do not certify anyone with it, because certifying is what a buyer wants to wave at a board. A small firm has no such board, and we will not pretend the check is a seal.

Gartner said in February 2025 that 63% of respondents do not have, or are unsure they have, the data practices AI needs.³ Through 2026, companies will abandon 60% of AI projects that lack AI-ready data. A framework that scores data as running while the facts sit in a shared inbox is scoring a wish. We keep that idea in the room during the talk. The research is about the people Gartner asked and not about your shop. It still describes the failure we watch, where the project looks alive until the record has to carry it.

Cisco’s 2025 index surveyed 8,000 senior leaders at companies with more than 500 employees.⁴ The leaders it called pacesetters were near 13%, and their companies were 4 times more likely to move a pilot into production. That group is not a small business. The pattern still matters, because production follows a written process and not a demo. If you quote the 13% at a bakery and call the baker not a pacesetter, you have changed the group. The framework helps us avoid that slide. We score the bakery in words, and we do not borrow a big-company league table.

Score the four groups in words first. Then run the check if you want the band. Words come before numbers because a number is easy to flatter. If you jump to the band, you will argue with the number and not with the inbox.

Here is how we pick the words, so they do not turn into moods. Data is absent when nobody can name the record. It is partial when a record exists, but people keep a side copy they trust more. It is in place when one current record is the one a new person would open, and a system could read it without anyone retyping it. It is running when that record already feeds a job you do every week, and someone notices when a field is wrong.

Process is absent when the steps live in one person’s head. It is partial when the usual path is written down and the exceptions are folklore. It is in place when the exceptions sit on the same page and one person owns the handoff, even on the day they are out. It is running when everyone follows that page without a meeting to find it again.

People is absent when nobody has said who may send output to a customer, and nobody has noticed that someone already does. It is partial when use is casual and nobody knows who can say no. It is in place when a named person may use the tool, a named person may send, and a named person may stop it. It is running when that stop has been used at least once and not just promised.

The decision is absent when success means less chaos. It is partial when success has a name but nobody can end the trial. It is in place when success, the person who can stop it, and the rule for wrong output are all written as sentences. It is running when you have already stopped something because the rule said to. A decision that has never met a bad week is still a draft.

The band sits on top of those words and does not replace them. Early, Forming, Ready, and Running tell you the shape, and the four words tell you where to work. We have watched owners cheer for Ready and ignore a data score that was still partial. The framework makes that cheer walk past the weak group first.

A worked example

Say we are in a small bakery after the morning rush. One person runs the front, and another runs the kitchen and the special-order book. This is an example we walk through, and it is not an audit. We did not measure their flour, and we will not invent a before and after.

The owner wants a tool that answers catering emails and suggests a quote. The demo replied faster than the front lead can. The framework is the part where we ignore the speed and score the job.

First, data. The special-order book is a paper notebook in the kitchen. The front lead also keeps notes in the shop inbox, because customers write to the shop address. The owner has a price sheet that is mostly current. We ask which record a quote must match, and the three of them do not agree. So data is absent for the catering job, even though a notebook exists. A notebook that nobody treats as final is not the one place the facts live. Calling it in place because paper feels solid would be the wish Gartner warns about, where the project is dropped later because the data was never ready. We say absent.

Next, process. The usual path is real. An email comes in, the kitchen lead checks the book, the owner sets a price, and the front lead replies. The exception is a customer who changes the head count the day before, or a date the kitchen lead half promised on the phone. Nobody wrote that down, so process is partial. The usual path is not the whole process. If we scored it in place, the tool would invent the head count change and sound sure about it.

Then people. The front lead already drafts replies in a personal account when she falls behind, and then she pastes them into the inbox. The owner did not know that customer words were being drafted somewhere else. Nobody is named as the person who may send, and nobody is named as the person who may block a send. People is partial, and it leans toward absent on the permission that matters. We do not call it good enough just because the front lead is capable. Capable is not the same as allowed.

Last, the decision. The owner says success is more catering without more evenings. That is almost a sentence. He cannot say what would make him turn the tool off, except if it felt wrong. A feeling is not a rule, so the decision is partial. We will not average these four words into a cheerful Ready. Absent data and partial everything else is an Early or Forming shape. We would rather say that than borrow the pacesetter story for a bakery that was not in the sample.

We leave the owner with four words and one job, and not with a percentage or a certificate. The next move is to pick one record, either the notebook or the inbox, and make it the one that counts. Then write down the head count exception, and name who may send. After that, a band means something.

How we’d run it

First, we pick the job the tool is meant to touch. At the bakery it is the catering reply, from the first email to a price the kitchen can keep. If the owner says the framework should cover the whole shop, we narrow it. A framework about everything scores nothing.

Second, we put the front lead and the kitchen lead in the room, and not only the owner. The owner describes the price list. The front describes the inbox, and the kitchen describes the book. The framework is a pointing exercise, and people who do not touch the job cannot point.

Third, we give data, then process, then people, then the decision one of the four words. We say each word out loud and wait to see if anyone objects. An objection usually means we were about to flatter the group. We move a word down when we cannot point at the proof. We never move it up because the demo was good.

Fourth, we look at the check only then, if they want the band from 0 to 100. The band has to fit the four words. If data is absent, a low band should not surprise anyone, and we should not argue it upward to protect the owner’s mood.

Fifth, we choose one group to move up a single step. Absent data becomes partial when one record gets a name, and not when a tool arrives. Partial process becomes in place when the head count exception sits on the same page as the usual path. We do not try to move all four groups in one panicked week. We take one step with one owner, and then we look again.

Sixth, we keep Cisco and Gartner in their proper jobs. They explain why data and production matter. They do not grade the bakery. If someone in the room quotes a big-company index as the owner’s score, we put the framework sheet back in the middle of the table.

Common mistakes

The mistake we see most is using the strong word because it sounds like progress. Calling a shared inbox running is the mistake we named above, and it is common because the inbox is busy. Busy is not the same as a record a system can read. A busy inbox is just a pile with a pulse.

Another mistake is scoring the owner’s hopes and not the job. The owner wants catering to grow, but hope is not one of the four groups. The groups are data, process, people, and the decision. If those are weak, the hope is a wish with a login.

A third mistake is bringing a pacesetter comparison into a small firm. Cisco’s 2025 index covers senior leaders at companies with more than 500 employees. The pacesetters near 13%, and the 4 times better odds of moving a pilot into production, describe that group. The bakery is not in it. We keep the pattern, which is a written process, and we drop the ranking.

We also see people skip the decision because it feels soft. It is the group that keeps a trial from turning into a plan you are afraid to cancel. If nobody can stop the experiment, you do not have a framework. You have a purchase.

What we’d do

We would score in words and stay there until the words stop embarrassing us. In our view the band is a convenience and the four words are the work. A bakery that can say data is absent, and the kitchen lead’s notebook is what we will make true, is ahead of a bakery that can quote a readiness percentage it did not earn.

If we were the owner, we would stop answering catering emails from side drafts until the inbox and the book agree on which one wins. We would also write the head count change on the same card as the quote steps. Then we would take the check. If the band says Early, we would believe it and keep going, because the framework did its job when it stopped a fast yes.